Dong Flash v2.3

The small, fast one.

fictional fastchat

Dong Flash is for short answers at high volume: chat widgets, classification, and any job where waiting costs more than being slightly wrong. In testing on a GTX 750 it produced a token every 40 seconds.* Faster hardware is planned.

*Testing consisted of one prompt. The prompt was "hello".

Model id: abliterate-dong-flash-v2.3
Context: 32K
Input rate, per 1M tokens: $0.10
Output rate, per 1M tokens: $0.30

Planned. This endpoint does not exist yet.

curl

curl https://api.abliterate.app/v1/chat/completions \
  -H "Authorization: Bearer ab-demo-not-a-real-key" \
  -H "Content-Type: application/json" \
  -d '{"model": "abliterate-dong-flash-v2.3", "messages": [{"role": "user", "content": "Say hello."}]}'